FairGov Codebook

Pre-treatment survey

Metadata

Description

Dataset name: .

The dataset has N=207 rows and 35 columns. 0 rows have no missing values on any column.

Metadata for search engines

• Date published: 2022-11-16

x sikker_fg fg_kunnskap_1 fg_kunnskap_2 fg_kunnskap_3 fg_forbedring fg_likhet_1 fg_likhet_2 fg_likhet_3 fg_kunnskap2 fg_auto_flykt fg_region fg_kjennskap_nav fg_auto_nav fg_pari_nav fg_kjennskap_prv fg_auto_prv fg_imp_1 fg_imp_2 fg_imp_3 fg_flykt_post fg_nav_post fg_prv_post fg_tradeoff_1 fg_tradeoff_2 fg_tradeoff_3 demo_edu demo_party demo_trust demo_lr_scale post fg_sikkert_post fo_delib fo_media on_sikkert_post fg

#Variables

sikker_fg

Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.

Distribution

## No non-missing values to show.

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
sikker_fg	Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.	haven_labelled	207	0	Inf	NA	-Inf	2		F40.0	5

Value labels

Response choices

name	value
Jeg vil delta i undersøkelsen	1
Jeg samtykker ikke og vil ikke delta i undersøkelsen eller det påfølgende arrangementet	2

fg_kunnskap_1

First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence

Distribution

Distribution of values for fg_kunnskap_1

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap_1	First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence	haven_labelled	0	1	1	4	6	3.73913	0.7692925	6	▁▁▁▂▇▁▁▁	F40.0	5

Value labels

Response choices

name	value
Very good knowledge	1
Good knowledge	2
Somwhat good knowledge	3
Little knowledge	4
No knowledge at all	5
Do not know	6

fg_kunnskap_2

First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway

Distribution

Distribution of values for fg_kunnskap_2

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap_2	First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway	haven_labelled	0	1	1	4	6	4.019324	0.9188498	6	▁▁▁▂▇▁▅▁	F40.0	5

Value labels

Response choices

name	value
Very good knowledge	1
Good knowledge	2
Somwhat good knowledge	3
Little knowledge	4
No knowledge at all	5
Do not know	6

fg_kunnskap_3

First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work

Distribution

Distribution of values for fg_kunnskap_3

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap_3	First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work	haven_labelled	0	1	1	4	6	3.961353	1.060525	6	▁▂▁▃▇▁▆▁	F40.0	5

Value labels

Response choices

name	value
Very good knowledge	1
Good knowledge	2
Somwhat good knowledge	3
Little knowledge	4
No knowledge at all	5
Do not know	6

fg_forbedring

Based on what you know, do you think the Norwegian authorities’ use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?

Distribution

Distribution of values for fg_forbedring

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_forbedring	Based on what you know, do you think the Norwegian authorities’ use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?	haven_labelled	0	1	1	3	9	4.091787	2.2396	8	▃▇▂▂▁▁▁▃	F40.0	5

Value labels

Response choices

name	value
Very strong improvement	1
Strong improvement	2
Some improvement	3
Neither improvement or worsening	4
Some worsening	5
Strong worsening	6
Very strong worsening	7
Do not know	9

fg_likhet_1

And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it

Distribution

Distribution of values for fg_likhet_1

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_likhet_1	And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it	haven_labelled	0	1	1	5	8	4.797101	2.001793	7	▁▂▇▆▆▃▂▆	F40.0	5

Value labels

Response choices

name	value
Very much more	1
More	2
Somwhat more	3
Neither more or less	4
Less	6
Very much less	7
Do not know	8

fg_likhet_2

And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector

Distribution

Distribution of values for fg_likhet_2

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_likhet_2	And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector	haven_labelled	0	1	1	3	8	3.652174	1.914376	8	▁▆▇▃▂▁▁▂	F40.0	5

Value labels

Response choices

name	value
Very much more	1
More	2
Somwhat more	3
Neither more or less	4
Somewhat less	5
Less	6
Very much less	7
Do not know	8

fg_likhet_3

And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities’ decisions

Distribution

Distribution of values for fg_likhet_3

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_likhet_3	And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities’ decisions	haven_labelled	0	1	1	4	8	4.729469	1.763781	8	▁▁▃▇▃▂▁▃	F40.0	5

Value labels

Response choices

name	value
Very much more	1
More	2
Somwhat more	3
Neither more or less	4
Somewhat less	5
Less	6
Very much less	7
Do not know	8

fg_kunnskap2

Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?

Distribution

Distribution of values for fg_kunnskap2

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap2	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?	haven_labelled	0	1	1	4	8	4.36715	1.376108	8	▁▂▅▇▇▂▁▁	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_auto_flykt

Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?

Distribution

Distribution of values for fg_auto_flykt

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_auto_flykt	Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	0	1	1	1	4	1.328502	0.7747904	3	▇▁▂▁▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	4

fg_region

Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?

Distribution

Distribution of values for fg_region

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_region	Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?	haven_labelled	0	1	1	4	8	4.560386	1.531221	8	▁▂▃▇▇▃▁▂	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_kjennskap_nav

Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?

Distribution

Distribution of values for fg_kjennskap_nav

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kjennskap_nav	Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?	haven_labelled	0	1	1	5	8	4.381642	1.658825	8	▁▂▅▃▇▂▁▁	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_auto_nav

Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?

Distribution

Distribution of values for fg_auto_nav

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_auto_nav	Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	0	1	1	1	3	1.246377	0.5234074	3	▇▁▁▂▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_pari_nav

Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?

Distribution

Distribution of values for fg_pari_nav

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_pari_nav	Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?	haven_labelled	0	1	1	1	3	1.603865	0.735761	3	▇▁▁▅▁▁▁▂	F40.0	5

Value labels

Response choices

name	value
The first algorithm seems most fair	1
The second algorithm seems most fair	2
Do not know	3

fg_kjennskap_prv

Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?

Distribution

Distribution of values for fg_kjennskap_prv

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kjennskap_prv	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?	haven_labelled	0	1	1	3	8	3.695652	1.784271	8	▃▅▇▃▇▂▁▁	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_auto_prv

Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?

Distribution

Distribution of values for fg_auto_prv

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_auto_prv	Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	0	1	1	1	3	1.120773	0.4061462	3	▇▁▁▁▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_imp_1

How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees

Distribution

Distribution of values for fg_imp_1

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_imp_1	How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees	haven_labelled	0	1	2	4	6	4.004831	0.766391	6	▁▃▁▇▁▃▁▁	F40.0	5

Value labels

Response choices

name	value
Not important at all	1
Not important	2
Somwhat important	3
Very important	4
Extremely improtant	5
Do not know	6

fg_imp_2

How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting

Distribution

Distribution of values for fg_imp_2

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_imp_2	How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting	haven_labelled	0	1	1	4	6	3.869565	0.8345636	6	▁▁▁▃▇▁▂▁	F40.0	5

Value labels

Response choices

name	value
Not important at all	1
Not important	2
Somwhat important	3
Very important	4
Extremely improtant	5
Do not know	6

fg_imp_3

How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates

Distribution

Distribution of values for fg_imp_3

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_imp_3	How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates	haven_labelled	0	1	2	5	6	4.434783	0.7532118	6	▁▂▁▅▁▇▁▁	F40.0	5

Value labels

Response choices

name	value
Not important at all	1
Not important	2
Somwhat important	3
Very important	4
Extremely improtant	5
Do not know	6

fg_flykt_post

Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?

Distribution

Distribution of values for fg_flykt_post

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_flykt_post	Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?	haven_labelled	0	1	1	2	3	1.719807	0.5385561	3	▅▁▁▇▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_nav_post

Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?

Distribution

Distribution of values for fg_nav_post

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_nav_post	Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?	haven_labelled	0	1	1	2	3	1.594203	0.5990002	3	▇▁▁▇▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_prv_post

Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?

Distribution

Distribution of values for fg_prv_post

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_prv_post	Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?	haven_labelled	0	1	1	2	3	1.652174	0.6860486	3	▇▁▁▇▁▁▁▂	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_tradeoff_1

Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?

Distribution

Distribution of values for fg_tradeoff_1

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_tradeoff_1	Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?	haven_labelled	0	1	1	1	3	1.657005	0.7840873	3	▇▁▁▅▁▁▁▃	F40.0	5

Value labels

Response choices

name	value
A	1
B	2
Do not know	3

fg_tradeoff_2

Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two

Distribution

Distribution of values for fg_tradeoff_2

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_tradeoff_2	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two	haven_labelled	0	1	1	2	3	1.975845	0.8614709	3	▇▁▁▆▁▁▁▇	F40.0	5

Value labels

Response choices

name	value
X	1
Y	2
Do not know	3

fg_tradeoff_3

Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive

Distribution

Distribution of values for fg_tradeoff_3

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_tradeoff_3	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive	haven_labelled	1	0.9951691	1	2	3	1.985437	0.7869556	3	▆▁▁▇▁▁▁▆	F40.0	5

Value labels

Response choices

name	value
One	1
Two	2
Do not know	3

demo_edu

What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.

Distribution

Distribution of values for demo_edu

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
demo_edu	What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.	haven_labelled	0	1	2	6	9	5.560386	1.571894	7	▁▅▁▃▇▇▁▁	F40.0	5

Value labels

Response choices

name	value
No eductation	1
Pre-school	2
High school	3
Vocational school level (includes educations that are based on upper secondary school, but which are not approved as university and college education)	5
University or collage, up to 4 yearrs	6
University or collage, more than 4 yearrs	7
Other	9

demo_party

Which party would you vote for if there were a parliamentary election tomorrow?

Distribution

Distribution of values for demo_party

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
demo_party	Which party would you vote for if there were a parliamentary election tomorrow?	haven_labelled	1	0.9951691	1	6	13	6.145631	3.683603	12	▇▃▁▇▃▂▂▃	F40.0	5

Value labels

Response choices

name	value
Progress party	1
Conservative party	2
Liberal Party	3
Christian democrats	4
Green party	13
Center party	5
Labour party	6
Socialist party	7
Red party	8
Other	9
Could not vote	10
Would not vote	11

demo_trust

How much trust or distrust do you have in scientists?

Distribution

Distribution of values for demo_trust

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
demo_trust	How much trust or distrust do you have in scientists?	haven_labelled	0	1	1	2	7	2.386473	1.072793	8	▂▇▂▂▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Very high trust	1
High trust	2
Some trust	3
Neither trust nor mistrust	4
Some mistrust	5
High mistrust	6
Very high mistrust	7
Do not know	8

demo_lr_scale

In politics, you often talk about the ‘left’ and the ‘right’. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?

Distribution

Distribution of values for demo_lr_scale

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
demo_lr_scale	In politics, you often talk about the ‘left’ and the ‘right’. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?	haven_labelled	1	0.9951691	0	5	11	5.165049	2.79755	12	▂▃▇▆▃▆▁▃	F40.0	5

Value labels

Response choices

name	value
0 - Left	0
1	1
2	2
3	3
4	4
5	5
6	6
7	7
8	8
9	9
10	10
11- Right	11

post

0 == Pre deliberation survey responses, 1 == post delibiration survey

Distribution

Distribution of values for post

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	hist
post	0 == Pre deliberation survey responses, 1 == post delibiration survey	numeric	0	1	0	0	0	0	0	▁▁▇▁▁

fg_sikkert_post

Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?

Distribution

## No non-missing values to show.

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
fg_sikkert_post	Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?	haven_labelled	207	0	Inf	NA	-Inf	3		F40.0	5

Value labels

Response choices

name	value
Veldig negativ	1
Veldig negativ	2
Veldig negativ	3

fo_delib

Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?

Distribution

## No non-missing values to show.

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
fo_delib	Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?	haven_labelled	207	0	Inf	NA	-Inf	4		F40.0	5

Value labels

Response choices

name	value
Veldig negativ	1
Veldig negativ	2
Veldig negativ	3
Veldig negativ	4

fo_media

Vi takker så mye for din deltakelse!

Helt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg?

Svaret ditt er ikke bindende.

Distribution

## No non-missing values to show.

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
fo_media	Vi takker så mye for din deltakelse!

Helt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg?

Svaret ditt er ikke bindende. |haven_labelled | 207| 0|Inf |NA |-Inf | 2| |F40.0 |5 |

Value labels

Response choices

name	value
Veldig negativ	4
Veldig negativ	5

on_sikkert_post

Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen?

Distribution

## No non-missing values to show.

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
on_sikkert_post	Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen?	haven_labelled	207	0	Inf	NA	-Inf	3		F40.0	5

Value labels

Response choices

name	value
Veldig negativ	1
Veldig negativ	2
Veldig negativ	3

fg

1 == Treatment group (AI), 0 == control group (carbon capture)

Distribution

Distribution of values for fg

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	hist
fg	1 == Treatment group (AI), 0 == control group (carbon capture)	numeric	0	1	0	1	1	0.5700483	0.4962691	▆▁▁▁▇

Missingness report

Codebook table

name	label	data_type	value_labels	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
sikker_fg	Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.	haven_labelled	1. Jeg vil delta i undersøkelsen, 2. Jeg samtykker ikke og vil ikke delta i undersøkelsen eller det påfølgende arrangementet	207	0.0000000	Inf	NA	-Inf	NaN	NA	2		F40.0	5
fg_kunnskap_1	First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence	haven_labelled	1. Very good knowledge, 2. Good knowledge, 3. Somwhat good knowledge, 4. Little knowledge, 5. No knowledge at all, 6. Do not know	0	1.0000000	1	4	6	3.7391304	0.7692925	6	▁▁▁▂▇▁▁▁	F40.0	5
fg_kunnskap_2	First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway	haven_labelled	1. Very good knowledge, 2. Good knowledge, 3. Somwhat good knowledge, 4. Little knowledge, 5. No knowledge at all, 6. Do not know	0	1.0000000	1	4	6	4.0193237	0.9188498	6	▁▁▁▂▇▁▅▁	F40.0	5
fg_kunnskap_3	First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work	haven_labelled	1. Very good knowledge, 2. Good knowledge, 3. Somwhat good knowledge, 4. Little knowledge, 5. No knowledge at all, 6. Do not know	0	1.0000000	1	4	6	3.9613527	1.0605247	6	▁▂▁▃▇▁▆▁	F40.0	5
fg_forbedring	Based on what you know, do you think the Norwegian authorities’ use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?	haven_labelled	1. Very strong improvement, 2. Strong improvement, 3. Some improvement, 4. Neither improvement or worsening, 5. Some worsening, 6. Strong worsening, 7. Very strong worsening, 9. Do not know	0	1.0000000	1	3	9	4.0917874	2.2395995	8	▃▇▂▂▁▁▁▃	F40.0	5
fg_likhet_1	And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it	haven_labelled	1. Very much more, 2. More, 3. Somwhat more, 4. Neither more or less, 6. Less, 7. Very much less, 8. Do not know	0	1.0000000	1	5	8	4.7971014	2.0017932	7	▁▂▇▆▆▃▂▆	F40.0	5
fg_likhet_2	And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector	haven_labelled	1. Very much more, 2. More, 3. Somwhat more, 4. Neither more or less, 5. Somewhat less, 6. Less, 7. Very much less, 8. Do not know	0	1.0000000	1	3	8	3.6521739	1.9143765	8	▁▆▇▃▂▁▁▂	F40.0	5
fg_likhet_3	And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities’ decisions	haven_labelled	1. Very much more, 2. More, 3. Somwhat more, 4. Neither more or less, 5. Somewhat less, 6. Less, 7. Very much less, 8. Do not know	0	1.0000000	1	4	8	4.7294686	1.7637810	8	▁▁▃▇▃▂▁▃	F40.0	5
fg_kunnskap2	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	0	1.0000000	1	4	8	4.3671498	1.3761079	8	▁▂▅▇▇▂▁▁	F40.0	5
fg_auto_flykt	Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 4. Do not know	0	1.0000000	1	1	4	1.3285024	0.7747904	3	▇▁▂▁▁▁▁▁	F40.0	5
fg_region	Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	0	1.0000000	1	4	8	4.5603865	1.5312207	8	▁▂▃▇▇▃▁▂	F40.0	5
fg_kjennskap_nav	Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	0	1.0000000	1	5	8	4.3816425	1.6588249	8	▁▂▅▃▇▂▁▁	F40.0	5
fg_auto_nav	Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	0	1.0000000	1	1	3	1.2463768	0.5234074	3	▇▁▁▂▁▁▁▁	F40.0	5
fg_pari_nav	Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?	haven_labelled	1. The first algorithm seems most fair, 2. The second algorithm seems most fair, 3. Do not know	0	1.0000000	1	1	3	1.6038647	0.7357610	3	▇▁▁▅▁▁▁▂	F40.0	5
fg_kjennskap_prv	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	0	1.0000000	1	3	8	3.6956522	1.7842707	8	▃▅▇▃▇▂▁▁	F40.0	5
fg_auto_prv	Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	0	1.0000000	1	1	3	1.1207729	0.4061462	3	▇▁▁▁▁▁▁▁	F40.0	5
fg_imp_1	How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees	haven_labelled	1. Not important at all, 2. Not important, 3. Somwhat important, 4. Very important, 5. Extremely improtant, 6. Do not know	0	1.0000000	2	4	6	4.0048309	0.7663910	6	▁▃▁▇▁▃▁▁	F40.0	5
fg_imp_2	How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting	haven_labelled	1. Not important at all, 2. Not important, 3. Somwhat important, 4. Very important, 5. Extremely improtant, 6. Do not know	0	1.0000000	1	4	6	3.8695652	0.8345636	6	▁▁▁▃▇▁▂▁	F40.0	5
fg_imp_3	How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates	haven_labelled	1. Not important at all, 2. Not important, 3. Somwhat important, 4. Very important, 5. Extremely improtant, 6. Do not know	0	1.0000000	2	5	6	4.4347826	0.7532118	6	▁▂▁▅▁▇▁▁	F40.0	5
fg_flykt_post	Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	0	1.0000000	1	2	3	1.7198068	0.5385561	3	▅▁▁▇▁▁▁▁	F40.0	5
fg_nav_post	Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	0	1.0000000	1	2	3	1.5942029	0.5990002	3	▇▁▁▇▁▁▁▁	F40.0	5
fg_prv_post	Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	0	1.0000000	1	2	3	1.6521739	0.6860486	3	▇▁▁▇▁▁▁▂	F40.0	5
fg_tradeoff_1	Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?	haven_labelled	1. A, 2. B, 3. Do not know	0	1.0000000	1	1	3	1.6570048	0.7840873	3	▇▁▁▅▁▁▁▃	F40.0	5
fg_tradeoff_2	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two	haven_labelled	1. X, 2. Y, 3. Do not know	0	1.0000000	1	2	3	1.9758454	0.8614709	3	▇▁▁▆▁▁▁▇	F40.0	5
fg_tradeoff_3	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive	haven_labelled	1. One, 2. Two, 3. Do not know	1	0.9951691	1	2	3	1.9854369	0.7869556	3	▆▁▁▇▁▁▁▆	F40.0	5
demo_edu	What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.	haven_labelled	1. No eductation, 2. Pre-school, 3. High school, 5. Vocational school level (includes educations that are based on upper secondary school, but which are not approved as university and college education), 6. University or collage, up to 4 yearrs, 7. University or collage, more than 4 yearrs, 9. Other	0	1.0000000	2	6	9	5.5603865	1.5718939	7	▁▅▁▃▇▇▁▁	F40.0	5
demo_party	Which party would you vote for if there were a parliamentary election tomorrow?	haven_labelled	1. Progress party, 2. Conservative party, 3. Liberal Party, 4. Christian democrats, 13. Green party, 5. Center party, 6. Labour party, 7. Socialist party, 8. Red party, 9. Other, 10. Could not vote, 11. Would not vote	1	0.9951691	1	6	13	6.1456311	3.6836031	12	▇▃▁▇▃▂▂▃	F40.0	5
demo_trust	How much trust or distrust do you have in scientists?	haven_labelled	1. Very high trust, 2. High trust, 3. Some trust, 4. Neither trust nor mistrust, 5. Some mistrust, 6. High mistrust, 7. Very high mistrust, 8. Do not know	0	1.0000000	1	2	7	2.3864734	1.0727927	8	▂▇▂▂▁▁▁▁	F40.0	5
demo_lr_scale	In politics, you often talk about the ‘left’ and the ‘right’. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?	haven_labelled	0. 0 - Left, 1. 1, 2. 2, 3. 3, 4. 4, 5. 5, 6. 6, 7. 7, 8. 8, 9. 9, 10. 10, 11. 11- Right	1	0.9951691	0	5	11	5.1650485	2.7975498	12	▂▃▇▆▃▆▁▃	F40.0	5
post	0 == Pre deliberation survey responses, 1 == post delibiration survey	numeric	.	0	1.0000000	0	0	0	0.0000000	0.0000000	NA	▁▁▇▁▁	NA	NA
fg_sikkert_post	Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?	haven_labelled	1. Veldig negativ, 2. Veldig negativ, 3. Veldig negativ	207	0.0000000	Inf	NA	-Inf	NaN	NA	3		F40.0	5
fo_delib	Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?	haven_labelled	1. Veldig negativ, 2. Veldig negativ, 3. Veldig negativ, 4. Veldig negativ	207	0.0000000	Inf	NA	-Inf	NaN	NA	4		F40.0	5
fo_media	Vi takker så mye for din deltakelse!

Helt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg?

Svaret ditt er ikke bindende. |haven_labelled |4. Veldig negativ,
5. Veldig negativ | 207| 0.0000000|Inf |NA |-Inf | NaN| NA| 2| |F40.0 |5 | |on_sikkert_post |Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen? |haven_labelled |1. Veldig negativ,
2. Veldig negativ,
3. Veldig negativ | 207| 0.0000000|Inf |NA |-Inf | NaN| NA| 3| |F40.0 |5 | |fg |1 == Treatment group (AI), 0 == control group (carbon capture) |numeric |. | 0| 1.0000000|0 |1 |1 | 0.5700483| 0.4962691| NA|▆▁▁▁▇ |NA |NA |

JSON-LD metadata

The following JSON-LD can be found by search engines, if you share this codebook publicly on the web.

{
  "name": ".",
  "datePublished": "2022-11-16",
  "description": "The dataset has N=207 rows and 35 columns.\n0 rows have no missing values on any column.\n\n\n## Table of variables\nThis table contains variable names, labels, and number of missing values.\nSee the complete codebook for more.\n\n[truncated]\n\n### Note\nThis dataset was automatically described using the [codebook R package](https://rubenarslan.github.io/codebook/) (version 0.9.2).",
  "keywords": ["sikker_fg", "fg_kunnskap_1", "fg_kunnskap_2", "fg_kunnskap_3", "fg_forbedring", "fg_likhet_1", "fg_likhet_2", "fg_likhet_3", "fg_kunnskap2", "fg_auto_flykt", "fg_region", "fg_kjennskap_nav", "fg_auto_nav", "fg_pari_nav", "fg_kjennskap_prv", "fg_auto_prv", "fg_imp_1", "fg_imp_2", "fg_imp_3", "fg_flykt_post", "fg_nav_post", "fg_prv_post", "fg_tradeoff_1", "fg_tradeoff_2", "fg_tradeoff_3", "demo_edu", "demo_party", "demo_trust", "demo_lr_scale", "post", "fg_sikkert_post", "fo_delib", "fo_media", "on_sikkert_post", "fg"],
  "@context": "http://schema.org/",
  "@type": "Dataset",
  "variableMeasured": [
    {
      "name": "sikker_fg",
      "description": "Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.",
      "value": "1. Jeg vil delta i undersøkelsen,\n2. Jeg samtykker ikke og vil ikke delta i undersøkelsen eller det påfølgende arrangementet",
      "maxValue": 2,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap_1",
      "description": "First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence",
      "value": "1. Very good knowledge,\n2. Good knowledge,\n3. Somwhat good knowledge,\n4. Little knowledge,\n5. No knowledge at all,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap_2",
      "description": "First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway",
      "value": "1. Very good knowledge,\n2. Good knowledge,\n3. Somwhat good knowledge,\n4. Little knowledge,\n5. No knowledge at all,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap_3",
      "description": "First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work",
      "value": "1. Very good knowledge,\n2. Good knowledge,\n3. Somwhat good knowledge,\n4. Little knowledge,\n5. No knowledge at all,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_forbedring",
      "description": "Based on what you know, do you think the Norwegian authorities' use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?",
      "value": "1. Very strong improvement,\n2. Strong improvement,\n3. Some improvement,\n4. Neither improvement or worsening,\n5. Some worsening,\n6. Strong worsening,\n7. Very strong worsening,\n9. Do not know",
      "maxValue": 9,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_likhet_1",
      "description": "And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it",
      "value": "1. Very much more,\n2. More,\n3. Somwhat more,\n4. Neither more or less,\n6. Less,\n7. Very much less,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_likhet_2",
      "description": "And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector",
      "value": "1. Very much more,\n2. More,\n3. Somwhat more,\n4. Neither more or less,\n5. Somewhat less,\n6. Less,\n7. Very much less,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_likhet_3",
      "description": "And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities' decisions",
      "value": "1. Very much more,\n2. More,\n3. Somwhat more,\n4. Neither more or less,\n5. Somewhat less,\n6. Less,\n7. Very much less,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap2",
      "description": "Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_auto_flykt",
      "description": "Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n4. Do not know",
      "maxValue": 4,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_region",
      "description": "Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kjennskap_nav",
      "description": "Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_auto_nav",
      "description": "Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_pari_nav",
      "description": "Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?",
      "value": "1. The first algorithm seems most fair,\n2. The second algorithm seems most fair,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kjennskap_prv",
      "description": "Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_auto_prv",
      "description": "Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_imp_1",
      "description": "How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees",
      "value": "1. Not important at all,\n2. Not important,\n3. Somwhat important,\n4. Very important,\n5. Extremely improtant,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_imp_2",
      "description": "How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting",
      "value": "1. Not important at all,\n2. Not important,\n3. Somwhat important,\n4. Very important,\n5. Extremely improtant,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_imp_3",
      "description": "How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates",
      "value": "1. Not important at all,\n2. Not important,\n3. Somwhat important,\n4. Very important,\n5. Extremely improtant,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_flykt_post",
      "description": "Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_nav_post",
      "description": "Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_prv_post",
      "description": "Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_tradeoff_1",
      "description": "Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?",
      "value": "1. A,\n2. B,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_tradeoff_2",
      "description": "Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two",
      "value": "1. X,\n2. Y,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_tradeoff_3",
      "description": "Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive",
      "value": "1. One,\n2. Two,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_edu",
      "description": "What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.",
      "value": "1. No eductation,\n2. Pre-school,\n3. High school,\n5. Vocational school level (includes educations that are based on upper secondary school, but which are not approved as university and college education),\n6. University or collage, up to 4 yearrs,\n7. University or collage, more than 4 yearrs,\n9. Other",
      "maxValue": 9,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_party",
      "description": "Which party would you vote for if there were a parliamentary election tomorrow?",
      "value": "1. Progress party,\n2. Conservative party,\n3. Liberal Party,\n4. Christian democrats,\n13. Green party,\n5. Center party,\n6. Labour party,\n7. Socialist party,\n8. Red party,\n9. Other,\n10. Could not vote,\n11. Would not vote",
      "maxValue": 13,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_trust",
      "description": "How much trust or distrust do you have in scientists?",
      "value": "1. Very high trust,\n2. High trust,\n3. Some trust,\n4. Neither trust nor mistrust,\n5. Some mistrust,\n6. High mistrust,\n7. Very high mistrust,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_lr_scale",
      "description": "In politics, you often talk about the 'left' and the 'right'. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?",
      "value": "0. 0 - Left,\n1. 1,\n2. 2,\n3. 3,\n4. 4,\n5. 5,\n6. 6,\n7. 7,\n8. 8,\n9. 9,\n10. 10,\n11. 11- Right",
      "maxValue": 11,
      "minValue": 0,
      "@type": "propertyValue"
    },
    {
      "name": "post",
      "description": "0 == Pre deliberation survey responses, 1 == post delibiration survey",
      "value": ". ",
      "maxValue": "-Inf",
      "minValue": "Inf",
      "@type": "propertyValue"
    },
    {
      "name": "fg_sikkert_post",
      "description": "Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?",
      "value": "1. Veldig negativ,\n2. Veldig negativ,\n3. Veldig negativ",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fo_delib",
      "description": "Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?",
      "value": "1. Veldig negativ,\n2. Veldig negativ,\n3. Veldig negativ,\n4. Veldig negativ",
      "maxValue": 4,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fo_media",
      "description": "Vi takker så mye for din deltakelse!  \r\n\r\nHelt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg? \r\n\r\nSvaret ditt er ikke bindende.",
      "value": "4. Veldig negativ,\n5. Veldig negativ",
      "maxValue": 5,
      "minValue": 4,
      "@type": "propertyValue"
    },
    {
      "name": "on_sikkert_post",
      "description": "Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen?",
      "value": "1. Veldig negativ,\n2. Veldig negativ,\n3. Veldig negativ",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg",
      "description": "1 == Treatment group (AI), 0 == control group (carbon capture)",
      "value": ". ",
      "maxValue": "-Inf",
      "minValue": "Inf",
      "@type": "propertyValue"
    }
  ]
}`

Post treatment survey

Metadata

Description

Dataset name: .

The dataset has N=207 rows and 35 columns. 0 rows have no missing values on any column.

Metadata for search engines

• Date published: 2022-11-16

x sikker_fg fg_kunnskap_1 fg_kunnskap_2 fg_kunnskap_3 fg_forbedring fg_likhet_1 fg_likhet_2 fg_likhet_3 fg_kunnskap2 fg_auto_flykt fg_region fg_kjennskap_nav fg_auto_nav fg_pari_nav fg_kjennskap_prv fg_auto_prv fg_imp_1 fg_imp_2 fg_imp_3 fg_flykt_post fg_nav_post fg_prv_post fg_tradeoff_1 fg_tradeoff_2 fg_tradeoff_3 demo_edu demo_party demo_trust demo_lr_scale post fg_sikkert_post fo_delib fo_media on_sikkert_post fg

#Variables

sikker_fg

Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.

Distribution

## No non-missing values to show.

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
sikker_fg	Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.	haven_labelled	207	0	Inf	NA	-Inf	2		F40.0	5

Value labels

Response choices

name	value
Jeg vil delta i undersøkelsen	1
Jeg samtykker ikke og vil ikke delta i undersøkelsen eller det påfølgende arrangementet	2

fg_kunnskap_1

First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence

Distribution

Distribution of values for fg_kunnskap_1

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap_1	First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence	haven_labelled	1	0.9951691	1	4	5	3.490291	0.7634344	6	▁▂▁▅▁▇▁▁	F40.0	5

Value labels

Response choices

name	value
Very good knowledge	1
Good knowledge	2
Somwhat good knowledge	3
Little knowledge	4
No knowledge at all	5
Do not know	6

fg_kunnskap_2

First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway

Distribution

Distribution of values for fg_kunnskap_2

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap_2	First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway	haven_labelled	1	0.9951691	1	4	6	3.747573	0.9649309	6	▁▂▁▂▇▁▃▁	F40.0	5

Value labels

Response choices

name	value
Very good knowledge	1
Good knowledge	2
Somwhat good knowledge	3
Little knowledge	4
No knowledge at all	5
Do not know	6

fg_kunnskap_3

First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work

Distribution

Distribution of values for fg_kunnskap_3

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap_3	First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work	haven_labelled	1	0.9951691	1	4	6	3.771845	0.9167192	6	▁▂▁▃▇▁▃▁	F40.0	5

Value labels

Response choices

name	value
Very good knowledge	1
Good knowledge	2
Somwhat good knowledge	3
Little knowledge	4
No knowledge at all	5
Do not know	6

fg_forbedring

Based on what you know, do you think the Norwegian authorities’ use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?

Distribution

Distribution of values for fg_forbedring

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_forbedring	Based on what you know, do you think the Norwegian authorities’ use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?	haven_labelled	1	0.9951691	1	3	9	3.339806	1.546378	8	▃▇▁▁▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Very strong improvement	1
Strong improvement	2
Some improvement	3
Neither improvement or worsening	4
Some worsening	5
Strong worsening	6
Very strong worsening	7
Do not know	9

fg_likhet_1

And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it

Distribution

Distribution of values for fg_likhet_1

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_likhet_1	And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it	haven_labelled	1	0.9951691	1	4	8	4.145631	1.847386	7	▁▃▇▅▅▃▂▂	F40.0	5

Value labels

Response choices

name	value
Very much more	1
More	2
Somwhat more	3
Neither more or less	4
Less	6
Very much less	7
Do not know	8

fg_likhet_2

And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector

Distribution

Distribution of values for fg_likhet_2

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_likhet_2	And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector	haven_labelled	1	0.9951691	1	3	8	3.019417	1.45153	8	▂▆▇▂▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Very much more	1
More	2
Somwhat more	3
Neither more or less	4
Somewhat less	5
Less	6
Very much less	7
Do not know	8

fg_likhet_3

And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities’ decisions

Distribution

Distribution of values for fg_likhet_3

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_likhet_3	And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities’ decisions	haven_labelled	1	0.9951691	2	4	8	4.330097	1.60723	8	▂▅▇▃▁▁▁▂	F40.0	5

Value labels

Response choices

name	value
Very much more	1
More	2
Somwhat more	3
Neither more or less	4
Somewhat less	5
Less	6
Very much less	7
Do not know	8

fg_kunnskap2

Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?

Distribution

Distribution of values for fg_kunnskap2

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap2	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?	haven_labelled	1	0.9951691	1	5	8	4.873786	1.307984	8	▁▁▂▂▇▅▁▁	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_auto_flykt

Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?

Distribution

Distribution of values for fg_auto_flykt

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_auto_flykt	Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1	0.9951691	1	1	4	1.160194	0.5022563	3	▇▁▁▁▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	4

fg_region

Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?

Distribution

Distribution of values for fg_region

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_region	Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?	haven_labelled	1	0.9951691	1	5	8	4.699029	1.563829	8	▁▁▅▇▇▆▁▂	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_kjennskap_nav

Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?

Distribution

Distribution of values for fg_kjennskap_nav

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kjennskap_nav	Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?	haven_labelled	1	0.9951691	1	5	8	4.834951	1.459015	8	▁▁▂▁▇▅▁▁	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_auto_nav

Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?

Distribution

Distribution of values for fg_auto_nav

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_auto_nav	Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1	0.9951691	1	1	3	1.145631	0.4050358	3	▇▁▁▁▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_pari_nav

Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?

Distribution

Distribution of values for fg_pari_nav

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_pari_nav	Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?	haven_labelled	1	0.9951691	1	1	3	1.587379	0.718831	3	▇▁▁▅▁▁▁▂	F40.0	5

Value labels

Response choices

name	value
The first algorithm seems most fair	1
The second algorithm seems most fair	2
Do not know	3

fg_kjennskap_prv

Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?

Distribution

Distribution of values for fg_kjennskap_prv

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kjennskap_prv	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?	haven_labelled	1	0.9951691	1	5	8	4.116505	1.645695	8	▂▃▅▂▇▃▁▁	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_auto_prv

Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?

Distribution

Distribution of values for fg_auto_prv

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_auto_prv	Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1	0.9951691	1	1	3	1.097087	0.3829173	3	▇▁▁▁▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_imp_1

How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees

Distribution

Distribution of values for fg_imp_1

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_imp_1	How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees	haven_labelled	1	0.9951691	1	4	6	3.927185	0.7774498	6	▁▁▁▃▇▁▃▁	F40.0	5

Value labels

Response choices

name	value
Not important at all	1
Not important	2
Somwhat important	3
Very important	4
Extremely improtant	5
Do not know	6

fg_imp_2

How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting

Distribution

Distribution of values for fg_imp_2

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_imp_2	How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting	haven_labelled	1	0.9951691	2	4	6	3.864078	0.7463543	6	▁▅▁▇▁▂▁▁	F40.0	5

Value labels

Response choices

name	value
Not important at all	1
Not important	2
Somwhat important	3
Very important	4
Extremely improtant	5
Do not know	6

fg_imp_3

How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates

Distribution

Distribution of values for fg_imp_3

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_imp_3	How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates	haven_labelled	1	0.9951691	1	5	6	4.446602	0.6946226	6	▁▁▁▁▆▁▇▁	F40.0	5

Value labels

Response choices

name	value
Not important at all	1
Not important	2
Somwhat important	3
Very important	4
Extremely improtant	5
Do not know	6

fg_flykt_post

Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?

Distribution

Distribution of values for fg_flykt_post

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_flykt_post	Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?	haven_labelled	1	0.9951691	1	2	3	1.718447	0.5745408	3	▅▁▁▇▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_nav_post

Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?

Distribution

Distribution of values for fg_nav_post

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_nav_post	Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?	haven_labelled	1	0.9951691	1	2	3	1.650485	0.5878234	3	▆▁▁▇▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_prv_post

Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?

Distribution

Distribution of values for fg_prv_post

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_prv_post	Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?	haven_labelled	1	0.9951691	1	2	3	1.597087	0.646032	3	▇▁▁▇▁▁▁▂	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_tradeoff_1

Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?

Distribution

Distribution of values for fg_tradeoff_1

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_tradeoff_1	Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?	haven_labelled	1	0.9951691	1	2	3	1.737864	0.8080228	3	▇▁▁▅▁▁▁▃	F40.0	5

Value labels

Response choices

name	value
A	1
B	2
Do not know	3

fg_tradeoff_2

Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two

Distribution

Distribution of values for fg_tradeoff_2

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_tradeoff_2	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two	haven_labelled	1	0.9951691	1	2	3	1.956311	0.8571274	3	▇▁▁▆▁▁▁▇	F40.0	5

Value labels

Response choices

name	value
X	1
Y	2
Do not know	3

fg_tradeoff_3

Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive

Distribution

Distribution of values for fg_tradeoff_3

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_tradeoff_3	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive	haven_labelled	1	0.9951691	1	2	3	1.975728	0.8111378	3	▇▁▁▇▁▁▁▇	F40.0	5

Value labels

Response choices

name	value
One	1
Two	2
Do not know	3

demo_edu

What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.

Distribution

## No non-missing values to show.

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
demo_edu	What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.	haven_labelled	207	0	Inf	NA	-Inf	7		F40.0	5

Value labels

Response choices

name	value
No eductation	1
Pre-school	2
High school	3
Vocational school level (includes educations that are based on upper secondary school, but which are not approved as university and college education)	5
University or collage, up to 4 yearrs	6
University or collage, more than 4 yearrs	7
Other	9

demo_party

Which party would you vote for if there were a parliamentary election tomorrow?

Distribution

## No non-missing values to show.

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
demo_party	Which party would you vote for if there were a parliamentary election tomorrow?	haven_labelled	207	0	Inf	NA	-Inf	12		F40.0	5

Value labels

Response choices

name	value
Progress party	1
Conservative party	2
Liberal Party	3
Christian democrats	4
Green party	13
Center party	5
Labour party	6
Socialist party	7
Red party	8
Other	9
Could not vote	10
Would not vote	11

demo_trust

How much trust or distrust do you have in scientists?

Distribution

## No non-missing values to show.

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
demo_trust	How much trust or distrust do you have in scientists?	haven_labelled	207	0	Inf	NA	-Inf	8		F40.0	5

Value labels

Response choices

name	value
Very high trust	1
High trust	2
Some trust	3
Neither trust nor mistrust	4
Some mistrust	5
High mistrust	6
Very high mistrust	7
Do not know	8

demo_lr_scale

In politics, you often talk about the ‘left’ and the ‘right’. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?

Distribution

## No non-missing values to show.

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
demo_lr_scale	In politics, you often talk about the ‘left’ and the ‘right’. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?	haven_labelled	207	0	Inf	NA	-Inf	12		F40.0	5

Value labels

Response choices

name	value
0 - Left	0
1	1
2	2
3	3
4	4
5	5
6	6
7	7
8	8
9	9
10	10
11- Right	11

post

0 == Pre deliberation survey responses, 1 == post delibiration survey

Distribution

Distribution of values for post

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	hist
post	0 == Pre deliberation survey responses, 1 == post delibiration survey	numeric	0	1	1	1	1	1	0	▁▁▇▁▁

fg_sikkert_post

Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?

Distribution

Distribution of values for fg_sikkert_post

90 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_sikkert_post	Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?	haven_labelled	90	0.5652174	1	3	3	2.666667	0.5085476	3	▁▁▁▃▁▁▁▇	F40.0	5

Value labels

Response choices

name	value
Veldig negativ	1
Veldig negativ	2
Veldig negativ	3

fo_delib

Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?

Distribution

Distribution of values for fo_delib

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fo_delib	Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?	haven_labelled	1	0.9951691	1	2	3	1.626214	0.6097518	4	▇▁▁▇▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Veldig negativ	1
Veldig negativ	2
Veldig negativ	3
Veldig negativ	4

fo_media

Vi takker så mye for din deltakelse!

Helt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg?

Svaret ditt er ikke bindende.

Distribution

Distribution of values for fo_media

2 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fo_media	Vi takker så mye for din deltakelse!

Helt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg?

Svaret ditt er ikke bindende. |haven_labelled | 2| 0.9903382|4 |4 |5 | 4.346342| 0.4769683| 2|▇▁▁▁▁▁▁▅ |F40.0 |5 |

Value labels

Response choices

name	value
Veldig negativ	4
Veldig negativ	5

on_sikkert_post

Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen?

Distribution

Distribution of values for on_sikkert_post

118 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
on_sikkert_post	Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen?	haven_labelled	118	0.4299517	1	3	3	2.52809	0.6757132	3	▁▁▁▃▁▁▁▇	F40.0	5

Value labels

Response choices

name	value
Veldig negativ	1
Veldig negativ	2
Veldig negativ	3

fg

1 == Treatment group (AI), 0 == control group (carbon capture)

Distribution

Distribution of values for fg

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	hist
fg	1 == Treatment group (AI), 0 == control group (carbon capture)	numeric	0	1	0	1	1	0.5700483	0.4962691	▆▁▁▁▇

Missingness report

Codebook table

name	label	data_type	value_labels	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
sikker_fg	Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.	haven_labelled	1. Jeg vil delta i undersøkelsen, 2. Jeg samtykker ikke og vil ikke delta i undersøkelsen eller det påfølgende arrangementet	207	0.0000000	Inf	NA	-Inf	NaN	NA	2		F40.0	5
fg_kunnskap_1	First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence	haven_labelled	1. Very good knowledge, 2. Good knowledge, 3. Somwhat good knowledge, 4. Little knowledge, 5. No knowledge at all, 6. Do not know	1	0.9951691	1	4	5	3.4902913	0.7634344	6	▁▂▁▅▁▇▁▁	F40.0	5
fg_kunnskap_2	First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway	haven_labelled	1. Very good knowledge, 2. Good knowledge, 3. Somwhat good knowledge, 4. Little knowledge, 5. No knowledge at all, 6. Do not know	1	0.9951691	1	4	6	3.7475728	0.9649309	6	▁▂▁▂▇▁▃▁	F40.0	5
fg_kunnskap_3	First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work	haven_labelled	1. Very good knowledge, 2. Good knowledge, 3. Somwhat good knowledge, 4. Little knowledge, 5. No knowledge at all, 6. Do not know	1	0.9951691	1	4	6	3.7718447	0.9167192	6	▁▂▁▃▇▁▃▁	F40.0	5
fg_forbedring	Based on what you know, do you think the Norwegian authorities’ use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?	haven_labelled	1. Very strong improvement, 2. Strong improvement, 3. Some improvement, 4. Neither improvement or worsening, 5. Some worsening, 6. Strong worsening, 7. Very strong worsening, 9. Do not know	1	0.9951691	1	3	9	3.3398058	1.5463783	8	▃▇▁▁▁▁▁▁	F40.0	5
fg_likhet_1	And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it	haven_labelled	1. Very much more, 2. More, 3. Somwhat more, 4. Neither more or less, 6. Less, 7. Very much less, 8. Do not know	1	0.9951691	1	4	8	4.1456311	1.8473858	7	▁▃▇▅▅▃▂▂	F40.0	5
fg_likhet_2	And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector	haven_labelled	1. Very much more, 2. More, 3. Somwhat more, 4. Neither more or less, 5. Somewhat less, 6. Less, 7. Very much less, 8. Do not know	1	0.9951691	1	3	8	3.0194175	1.4515296	8	▂▆▇▂▁▁▁▁	F40.0	5
fg_likhet_3	And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities’ decisions	haven_labelled	1. Very much more, 2. More, 3. Somwhat more, 4. Neither more or less, 5. Somewhat less, 6. Less, 7. Very much less, 8. Do not know	1	0.9951691	2	4	8	4.3300971	1.6072297	8	▂▅▇▃▁▁▁▂	F40.0	5
fg_kunnskap2	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	1	0.9951691	1	5	8	4.8737864	1.3079838	8	▁▁▂▂▇▅▁▁	F40.0	5
fg_auto_flykt	Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 4. Do not know	1	0.9951691	1	1	4	1.1601942	0.5022563	3	▇▁▁▁▁▁▁▁	F40.0	5
fg_region	Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	1	0.9951691	1	5	8	4.6990291	1.5638286	8	▁▁▅▇▇▆▁▂	F40.0	5
fg_kjennskap_nav	Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	1	0.9951691	1	5	8	4.8349515	1.4590146	8	▁▁▂▁▇▅▁▁	F40.0	5
fg_auto_nav	Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	1	0.9951691	1	1	3	1.1456311	0.4050358	3	▇▁▁▁▁▁▁▁	F40.0	5
fg_pari_nav	Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?	haven_labelled	1. The first algorithm seems most fair, 2. The second algorithm seems most fair, 3. Do not know	1	0.9951691	1	1	3	1.5873786	0.7188310	3	▇▁▁▅▁▁▁▂	F40.0	5
fg_kjennskap_prv	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	1	0.9951691	1	5	8	4.1165049	1.6456949	8	▂▃▅▂▇▃▁▁	F40.0	5
fg_auto_prv	Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	1	0.9951691	1	1	3	1.0970874	0.3829173	3	▇▁▁▁▁▁▁▁	F40.0	5
fg_imp_1	How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees	haven_labelled	1. Not important at all, 2. Not important, 3. Somwhat important, 4. Very important, 5. Extremely improtant, 6. Do not know	1	0.9951691	1	4	6	3.9271845	0.7774498	6	▁▁▁▃▇▁▃▁	F40.0	5
fg_imp_2	How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting	haven_labelled	1. Not important at all, 2. Not important, 3. Somwhat important, 4. Very important, 5. Extremely improtant, 6. Do not know	1	0.9951691	2	4	6	3.8640777	0.7463543	6	▁▅▁▇▁▂▁▁	F40.0	5
fg_imp_3	How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates	haven_labelled	1. Not important at all, 2. Not important, 3. Somwhat important, 4. Very important, 5. Extremely improtant, 6. Do not know	1	0.9951691	1	5	6	4.4466019	0.6946226	6	▁▁▁▁▆▁▇▁	F40.0	5
fg_flykt_post	Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	1	0.9951691	1	2	3	1.7184466	0.5745408	3	▅▁▁▇▁▁▁▁	F40.0	5
fg_nav_post	Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	1	0.9951691	1	2	3	1.6504854	0.5878234	3	▆▁▁▇▁▁▁▁	F40.0	5
fg_prv_post	Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	1	0.9951691	1	2	3	1.5970874	0.6460320	3	▇▁▁▇▁▁▁▂	F40.0	5
fg_tradeoff_1	Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?	haven_labelled	1. A, 2. B, 3. Do not know	1	0.9951691	1	2	3	1.7378641	0.8080228	3	▇▁▁▅▁▁▁▃	F40.0	5
fg_tradeoff_2	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two	haven_labelled	1. X, 2. Y, 3. Do not know	1	0.9951691	1	2	3	1.9563107	0.8571274	3	▇▁▁▆▁▁▁▇	F40.0	5
fg_tradeoff_3	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive	haven_labelled	1. One, 2. Two, 3. Do not know	1	0.9951691	1	2	3	1.9757282	0.8111378	3	▇▁▁▇▁▁▁▇	F40.0	5
demo_edu	What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.	haven_labelled	1. No eductation, 2. Pre-school, 3. High school, 5. Vocational school level (includes educations that are based on upper secondary school, but which are not approved as university and college education), 6. University or collage, up to 4 yearrs, 7. University or collage, more than 4 yearrs, 9. Other	207	0.0000000	Inf	NA	-Inf	NaN	NA	7		F40.0	5
demo_party	Which party would you vote for if there were a parliamentary election tomorrow?	haven_labelled	1. Progress party, 2. Conservative party, 3. Liberal Party, 4. Christian democrats, 13. Green party, 5. Center party, 6. Labour party, 7. Socialist party, 8. Red party, 9. Other, 10. Could not vote, 11. Would not vote	207	0.0000000	Inf	NA	-Inf	NaN	NA	12		F40.0	5
demo_trust	How much trust or distrust do you have in scientists?	haven_labelled	1. Very high trust, 2. High trust, 3. Some trust, 4. Neither trust nor mistrust, 5. Some mistrust, 6. High mistrust, 7. Very high mistrust, 8. Do not know	207	0.0000000	Inf	NA	-Inf	NaN	NA	8		F40.0	5
demo_lr_scale	In politics, you often talk about the ‘left’ and the ‘right’. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?	haven_labelled	0. 0 - Left, 1. 1, 2. 2, 3. 3, 4. 4, 5. 5, 6. 6, 7. 7, 8. 8, 9. 9, 10. 10, 11. 11- Right	207	0.0000000	Inf	NA	-Inf	NaN	NA	12		F40.0	5
post	0 == Pre deliberation survey responses, 1 == post delibiration survey	numeric	.	0	1.0000000	1	1	1	1.0000000	0.0000000	NA	▁▁▇▁▁	NA	NA
fg_sikkert_post	Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?	haven_labelled	1. Veldig negativ, 2. Veldig negativ, 3. Veldig negativ	90	0.5652174	1	3	3	2.6666667	0.5085476	3	▁▁▁▃▁▁▁▇	F40.0	5
fo_delib	Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?	haven_labelled	1. Veldig negativ, 2. Veldig negativ, 3. Veldig negativ, 4. Veldig negativ	1	0.9951691	1	2	3	1.6262136	0.6097518	4	▇▁▁▇▁▁▁▁	F40.0	5
fo_media	Vi takker så mye for din deltakelse!

Helt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg?

Svaret ditt er ikke bindende. |haven_labelled |4. Veldig negativ,
5. Veldig negativ | 2| 0.9903382|4 |4 |5 | 4.3463415| 0.4769683| 2|▇▁▁▁▁▁▁▅ |F40.0 |5 | |on_sikkert_post |Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen? |haven_labelled |1. Veldig negativ,
2. Veldig negativ,
3. Veldig negativ | 118| 0.4299517|1 |3 |3 | 2.5280899| 0.6757132| 3|▁▁▁▃▁▁▁▇ |F40.0 |5 | |fg |1 == Treatment group (AI), 0 == control group (carbon capture) |numeric |. | 0| 1.0000000|0 |1 |1 | 0.5700483| 0.4962691| NA|▆▁▁▁▇ |NA |NA |

JSON-LD metadata

The following JSON-LD can be found by search engines, if you share this codebook publicly on the web.

{
  "name": ".",
  "datePublished": "2022-11-16",
  "description": "The dataset has N=207 rows and 35 columns.\n0 rows have no missing values on any column.\n\n\n## Table of variables\nThis table contains variable names, labels, and number of missing values.\nSee the complete codebook for more.\n\n[truncated]\n\n### Note\nThis dataset was automatically described using the [codebook R package](https://rubenarslan.github.io/codebook/) (version 0.9.2).",
  "keywords": ["sikker_fg", "fg_kunnskap_1", "fg_kunnskap_2", "fg_kunnskap_3", "fg_forbedring", "fg_likhet_1", "fg_likhet_2", "fg_likhet_3", "fg_kunnskap2", "fg_auto_flykt", "fg_region", "fg_kjennskap_nav", "fg_auto_nav", "fg_pari_nav", "fg_kjennskap_prv", "fg_auto_prv", "fg_imp_1", "fg_imp_2", "fg_imp_3", "fg_flykt_post", "fg_nav_post", "fg_prv_post", "fg_tradeoff_1", "fg_tradeoff_2", "fg_tradeoff_3", "demo_edu", "demo_party", "demo_trust", "demo_lr_scale", "post", "fg_sikkert_post", "fo_delib", "fo_media", "on_sikkert_post", "fg"],
  "@context": "http://schema.org/",
  "@type": "Dataset",
  "variableMeasured": [
    {
      "name": "sikker_fg",
      "description": "Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.",
      "value": "1. Jeg vil delta i undersøkelsen,\n2. Jeg samtykker ikke og vil ikke delta i undersøkelsen eller det påfølgende arrangementet",
      "maxValue": 2,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap_1",
      "description": "First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence",
      "value": "1. Very good knowledge,\n2. Good knowledge,\n3. Somwhat good knowledge,\n4. Little knowledge,\n5. No knowledge at all,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap_2",
      "description": "First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway",
      "value": "1. Very good knowledge,\n2. Good knowledge,\n3. Somwhat good knowledge,\n4. Little knowledge,\n5. No knowledge at all,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap_3",
      "description": "First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work",
      "value": "1. Very good knowledge,\n2. Good knowledge,\n3. Somwhat good knowledge,\n4. Little knowledge,\n5. No knowledge at all,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_forbedring",
      "description": "Based on what you know, do you think the Norwegian authorities' use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?",
      "value": "1. Very strong improvement,\n2. Strong improvement,\n3. Some improvement,\n4. Neither improvement or worsening,\n5. Some worsening,\n6. Strong worsening,\n7. Very strong worsening,\n9. Do not know",
      "maxValue": 9,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_likhet_1",
      "description": "And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it",
      "value": "1. Very much more,\n2. More,\n3. Somwhat more,\n4. Neither more or less,\n6. Less,\n7. Very much less,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_likhet_2",
      "description": "And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector",
      "value": "1. Very much more,\n2. More,\n3. Somwhat more,\n4. Neither more or less,\n5. Somewhat less,\n6. Less,\n7. Very much less,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_likhet_3",
      "description": "And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities' decisions",
      "value": "1. Very much more,\n2. More,\n3. Somwhat more,\n4. Neither more or less,\n5. Somewhat less,\n6. Less,\n7. Very much less,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap2",
      "description": "Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_auto_flykt",
      "description": "Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n4. Do not know",
      "maxValue": 4,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_region",
      "description": "Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kjennskap_nav",
      "description": "Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_auto_nav",
      "description": "Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_pari_nav",
      "description": "Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?",
      "value": "1. The first algorithm seems most fair,\n2. The second algorithm seems most fair,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kjennskap_prv",
      "description": "Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_auto_prv",
      "description": "Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_imp_1",
      "description": "How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees",
      "value": "1. Not important at all,\n2. Not important,\n3. Somwhat important,\n4. Very important,\n5. Extremely improtant,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_imp_2",
      "description": "How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting",
      "value": "1. Not important at all,\n2. Not important,\n3. Somwhat important,\n4. Very important,\n5. Extremely improtant,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_imp_3",
      "description": "How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates",
      "value": "1. Not important at all,\n2. Not important,\n3. Somwhat important,\n4. Very important,\n5. Extremely improtant,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_flykt_post",
      "description": "Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_nav_post",
      "description": "Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_prv_post",
      "description": "Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_tradeoff_1",
      "description": "Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?",
      "value": "1. A,\n2. B,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_tradeoff_2",
      "description": "Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two",
      "value": "1. X,\n2. Y,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_tradeoff_3",
      "description": "Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive",
      "value": "1. One,\n2. Two,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_edu",
      "description": "What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.",
      "value": "1. No eductation,\n2. Pre-school,\n3. High school,\n5. Vocational school level (includes educations that are based on upper secondary school, but which are not approved as university and college education),\n6. University or collage, up to 4 yearrs,\n7. University or collage, more than 4 yearrs,\n9. Other",
      "maxValue": 9,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_party",
      "description": "Which party would you vote for if there were a parliamentary election tomorrow?",
      "value": "1. Progress party,\n2. Conservative party,\n3. Liberal Party,\n4. Christian democrats,\n13. Green party,\n5. Center party,\n6. Labour party,\n7. Socialist party,\n8. Red party,\n9. Other,\n10. Could not vote,\n11. Would not vote",
      "maxValue": 13,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_trust",
      "description": "How much trust or distrust do you have in scientists?",
      "value": "1. Very high trust,\n2. High trust,\n3. Some trust,\n4. Neither trust nor mistrust,\n5. Some mistrust,\n6. High mistrust,\n7. Very high mistrust,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_lr_scale",
      "description": "In politics, you often talk about the 'left' and the 'right'. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?",
      "value": "0. 0 - Left,\n1. 1,\n2. 2,\n3. 3,\n4. 4,\n5. 5,\n6. 6,\n7. 7,\n8. 8,\n9. 9,\n10. 10,\n11. 11- Right",
      "maxValue": 11,
      "minValue": 0,
      "@type": "propertyValue"
    },
    {
      "name": "post",
      "description": "0 == Pre deliberation survey responses, 1 == post delibiration survey",
      "value": ". ",
      "maxValue": "-Inf",
      "minValue": "Inf",
      "@type": "propertyValue"
    },
    {
      "name": "fg_sikkert_post",
      "description": "Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?",
      "value": "1. Veldig negativ,\n2. Veldig negativ,\n3. Veldig negativ",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fo_delib",
      "description": "Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?",
      "value": "1. Veldig negativ,\n2. Veldig negativ,\n3. Veldig negativ,\n4. Veldig negativ",
      "maxValue": 4,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fo_media",
      "description": "Vi takker så mye for din deltakelse!  \r\n\r\nHelt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg? \r\n\r\nSvaret ditt er ikke bindende.",
      "value": "4. Veldig negativ,\n5. Veldig negativ",
      "maxValue": 5,
      "minValue": 4,
      "@type": "propertyValue"
    },
    {
      "name": "on_sikkert_post",
      "description": "Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen?",
      "value": "1. Veldig negativ,\n2. Veldig negativ,\n3. Veldig negativ",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg",
      "description": "1 == Treatment group (AI), 0 == control group (carbon capture)",
      "value": ". ",
      "maxValue": "-Inf",
      "minValue": "Inf",
      "@type": "propertyValue"
    }
  ]
}`

Combined

Metadata

Description

Dataset name: .

The dataset has N=414 rows and 35 columns. 0 rows have no missing values on any column.

Metadata for search engines

• Date published: 2022-11-16

x sikker_fg fg_kunnskap_1 fg_kunnskap_2 fg_kunnskap_3 fg_forbedring fg_likhet_1 fg_likhet_2 fg_likhet_3 fg_kunnskap2 fg_auto_flykt fg_region fg_kjennskap_nav fg_auto_nav fg_pari_nav fg_kjennskap_prv fg_auto_prv fg_imp_1 fg_imp_2 fg_imp_3 fg_flykt_post fg_nav_post fg_prv_post fg_tradeoff_1 fg_tradeoff_2 fg_tradeoff_3 demo_edu demo_party demo_trust demo_lr_scale post fg_sikkert_post fo_delib fo_media on_sikkert_post fg

#Variables

sikker_fg

Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.

Distribution

## No non-missing values to show.

414 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	n_value_labels	hist	format.spss	display_width
sikker_fg	Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.	haven_labelled	414	0	Inf	NA	-Inf	2		F40.0	5

Value labels

Response choices

name	value
Jeg vil delta i undersøkelsen	1
Jeg samtykker ikke og vil ikke delta i undersøkelsen eller det påfølgende arrangementet	2

fg_kunnskap_1

First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence

Distribution

Distribution of values for fg_kunnskap_1

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap_1	First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence	haven_labelled	1	0.9975845	1	4	6	3.615012	0.7755157	6	▁▁▁▃▇▁▁▁	F40.0	5

Value labels

Response choices

name	value
Very good knowledge	1
Good knowledge	2
Somwhat good knowledge	3
Little knowledge	4
No knowledge at all	5
Do not know	6

fg_kunnskap_2

First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway

Distribution

Distribution of values for fg_kunnskap_2

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap_2	First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway	haven_labelled	1	0.9975845	1	4	6	3.883777	0.9507551	6	▁▂▁▂▇▁▃▁	F40.0	5

Value labels

Response choices

name	value
Very good knowledge	1
Good knowledge	2
Somwhat good knowledge	3
Little knowledge	4
No knowledge at all	5
Do not know	6

fg_kunnskap_3

First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work

Distribution

Distribution of values for fg_kunnskap_3

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap_3	First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work	haven_labelled	1	0.9975845	1	4	6	3.866828	0.994738	6	▁▂▁▃▇▁▅▁	F40.0	5

Value labels

Response choices

name	value
Very good knowledge	1
Good knowledge	2
Somwhat good knowledge	3
Little knowledge	4
No knowledge at all	5
Do not know	6

fg_forbedring

Based on what you know, do you think the Norwegian authorities’ use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?

Distribution

Distribution of values for fg_forbedring

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_forbedring	Based on what you know, do you think the Norwegian authorities’ use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?	haven_labelled	1	0.9975845	1	3	9	3.716707	1.959452	8	▃▇▂▂▁▁▁▂	F40.0	5

Value labels

Response choices

name	value
Very strong improvement	1
Strong improvement	2
Some improvement	3
Neither improvement or worsening	4
Some worsening	5
Strong worsening	6
Very strong worsening	7
Do not know	9

fg_likhet_1

And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it

Distribution

Distribution of values for fg_likhet_1

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_likhet_1	And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it	haven_labelled	1	0.9975845	1	4	8	4.472155	1.951431	7	▁▃▇▅▅▃▂▃	F40.0	5

Value labels

Response choices

name	value
Very much more	1
More	2
Somwhat more	3
Neither more or less	4
Less	6
Very much less	7
Do not know	8

fg_likhet_2

And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector

Distribution

Distribution of values for fg_likhet_2

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_likhet_2	And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector	haven_labelled	1	0.9975845	1	3	8	3.336562	1.72659	8	▂▆▇▃▁▁▁▂	F40.0	5

Value labels

Response choices

name	value
Very much more	1
More	2
Somwhat more	3
Neither more or less	4
Somewhat less	5
Less	6
Very much less	7
Do not know	8

fg_likhet_3

And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities’ decisions

Distribution

Distribution of values for fg_likhet_3

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_likhet_3	And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities’ decisions	haven_labelled	1	0.9975845	1	4	8	4.530266	1.697279	8	▁▂▃▇▃▂▁▂	F40.0	5

Value labels

Response choices

name	value
Very much more	1
More	2
Somwhat more	3
Neither more or less	4
Somewhat less	5
Less	6
Very much less	7
Do not know	8

fg_kunnskap2

Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?

Distribution

Distribution of values for fg_kunnskap2

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kunnskap2	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?	haven_labelled	1	0.9975845	1	5	8	4.619855	1.364705	8	▁▁▃▅▇▃▁▁	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_auto_flykt

Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?

Distribution

Distribution of values for fg_auto_flykt

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_auto_flykt	Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1	0.9975845	1	1	4	1.244552	0.6578507	3	▇▁▁▁▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	4

fg_region

Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?

Distribution

Distribution of values for fg_region

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_region	Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?	haven_labelled	1	0.9975845	1	5	8	4.62954	1.547249	8	▁▂▃▇▇▆▁▂	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_kjennskap_nav

Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?

Distribution

Distribution of values for fg_kjennskap_nav

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kjennskap_nav	Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?	haven_labelled	1	0.9975845	1	5	8	4.607748	1.576877	8	▁▂▃▂▇▃▁▁	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_auto_nav

Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?

Distribution

Distribution of values for fg_auto_nav

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_auto_nav	Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1	0.9975845	1	1	3	1.196126	0.4702659	3	▇▁▁▁▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_pari_nav

Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?

Distribution

Distribution of values for fg_pari_nav

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_pari_nav	Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?	haven_labelled	1	0.9975845	1	1	3	1.595642	0.7265294	3	▇▁▁▅▁▁▁▂	F40.0	5

Value labels

Response choices

name	value
The first algorithm seems most fair	1
The second algorithm seems most fair	2
Do not know	3

fg_kjennskap_prv

Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?

Distribution

Distribution of values for fg_kjennskap_prv

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_kjennskap_prv	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?	haven_labelled	1	0.9975845	1	4	8	3.905569	1.727362	8	▂▃▅▃▇▂▁▁	F40.0	5

Value labels

Response choices

name	value
Oppose very strongly	1
Oppose strongly	2
Oppose somewhat	3
Neither oppose or support	4
Support somewhat	5
Support strongly	6
Support very stringly	7
Do not know	8

fg_auto_prv

Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?

Distribution

Distribution of values for fg_auto_prv

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_auto_prv	Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1	0.9975845	1	1	3	1.108959	0.3944298	3	▇▁▁▁▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_imp_1

How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees

Distribution

Distribution of values for fg_imp_1

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_imp_1	How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees	haven_labelled	1	0.9975845	1	4	6	3.966102	0.7719686	6	▁▁▁▃▇▁▃▁	F40.0	5

Value labels

Response choices

name	value
Not important at all	1
Not important	2
Somwhat important	3
Very important	4
Extremely improtant	5
Do not know	6

fg_imp_2

How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting

Distribution

Distribution of values for fg_imp_2

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_imp_2	How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting	haven_labelled	1	0.9975845	1	4	6	3.866828	0.7908388	6	▁▁▁▅▇▁▂▁	F40.0	5

Value labels

Response choices

name	value
Not important at all	1
Not important	2
Somwhat important	3
Very important	4
Extremely improtant	5
Do not know	6

fg_imp_3

How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates

Distribution

Distribution of values for fg_imp_3

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_imp_3	How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates	haven_labelled	1	0.9975845	1	5	6	4.440678	0.7237252	6	▁▁▁▁▆▁▇▁	F40.0	5

Value labels

Response choices

name	value
Not important at all	1
Not important	2
Somwhat important	3
Very important	4
Extremely improtant	5
Do not know	6

fg_flykt_post

Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?

Distribution

Distribution of values for fg_flykt_post

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_flykt_post	Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?	haven_labelled	1	0.9975845	1	2	3	1.719128	0.5561197	3	▅▁▁▇▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_nav_post

Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?

Distribution

Distribution of values for fg_nav_post

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_nav_post	Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?	haven_labelled	1	0.9975845	1	2	3	1.622276	0.5934003	3	▇▁▁▇▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_prv_post

Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?

Distribution

Distribution of values for fg_prv_post

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_prv_post	Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?	haven_labelled	1	0.9975845	1	2	3	1.624697	0.6661513	3	▇▁▁▇▁▁▁▂	F40.0	5

Value labels

Response choices

name	value
Manual processing	1
Algorithmic processing	2
Do not know	3

fg_tradeoff_1

Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?

Distribution

Distribution of values for fg_tradeoff_1

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_tradeoff_1	Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?	haven_labelled	1	0.9975845	1	1	3	1.697337	0.7961788	3	▇▁▁▅▁▁▁▃	F40.0	5

Value labels

Response choices

name	value
A	1
B	2
Do not know	3

fg_tradeoff_2

Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two

Distribution

Distribution of values for fg_tradeoff_2

1 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_tradeoff_2	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two	haven_labelled	1	0.9975845	1	2	3	1.966102	0.8583194	3	▇▁▁▆▁▁▁▇	F40.0	5

Value labels

Response choices

name	value
X	1
Y	2
Do not know	3

fg_tradeoff_3

Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive

Distribution

Distribution of values for fg_tradeoff_3

2 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_tradeoff_3	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive	haven_labelled	2	0.9951691	1	2	3	1.980582	0.7981802	3	▇▁▁▇▁▁▁▇	F40.0	5

Value labels

Response choices

name	value
One	1
Two	2
Do not know	3

demo_edu

What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.

Distribution

Distribution of values for demo_edu

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
demo_edu	What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.	haven_labelled	207	0.5	2	6	9	5.560386	1.571894	7	▁▅▁▃▇▇▁▁	F40.0	5

Value labels

Response choices

name	value
No eductation	1
Pre-school	2
High school	3
Vocational school level (includes educations that are based on upper secondary school, but which are not approved as university and college education)	5
University or collage, up to 4 yearrs	6
University or collage, more than 4 yearrs	7
Other	9

demo_party

Which party would you vote for if there were a parliamentary election tomorrow?

Distribution

Distribution of values for demo_party

208 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
demo_party	Which party would you vote for if there were a parliamentary election tomorrow?	haven_labelled	208	0.4975845	1	6	13	6.145631	3.683603	12	▇▃▁▇▃▂▂▃	F40.0	5

Value labels

Response choices

name	value
Progress party	1
Conservative party	2
Liberal Party	3
Christian democrats	4
Green party	13
Center party	5
Labour party	6
Socialist party	7
Red party	8
Other	9
Could not vote	10
Would not vote	11

demo_trust

How much trust or distrust do you have in scientists?

Distribution

Distribution of values for demo_trust

207 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
demo_trust	How much trust or distrust do you have in scientists?	haven_labelled	207	0.5	1	2	7	2.386473	1.072793	8	▂▇▂▂▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Very high trust	1
High trust	2
Some trust	3
Neither trust nor mistrust	4
Some mistrust	5
High mistrust	6
Very high mistrust	7
Do not know	8

demo_lr_scale

In politics, you often talk about the ‘left’ and the ‘right’. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?

Distribution

Distribution of values for demo_lr_scale

208 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
demo_lr_scale	In politics, you often talk about the ‘left’ and the ‘right’. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?	haven_labelled	208	0.4975845	0	5	11	5.165049	2.79755	12	▂▃▇▆▃▆▁▃	F40.0	5

Value labels

Response choices

name	value
0 - Left	0
1	1
2	2
3	3
4	4
5	5
6	6
7	7
8	8
9	9
10	10
11- Right	11

post

0 == Pre deliberation survey responses, 1 == post delibiration survey

Distribution

Distribution of values for post

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	hist
post	0 == Pre deliberation survey responses, 1 == post delibiration survey	numeric	0	1	0	0.5	1	0.5	0.500605	▇▁▁▁▇

fg_sikkert_post

Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?

Distribution

Distribution of values for fg_sikkert_post

297 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fg_sikkert_post	Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?	haven_labelled	297	0.2826087	1	3	3	2.666667	0.5085476	3	▁▁▁▃▁▁▁▇	F40.0	5

Value labels

Response choices

name	value
Veldig negativ	1
Veldig negativ	2
Veldig negativ	3

fo_delib

Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?

Distribution

Distribution of values for fo_delib

208 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fo_delib	Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?	haven_labelled	208	0.4975845	1	2	3	1.626214	0.6097518	4	▇▁▁▇▁▁▁▁	F40.0	5

Value labels

Response choices

name	value
Veldig negativ	1
Veldig negativ	2
Veldig negativ	3
Veldig negativ	4

fo_media

Vi takker så mye for din deltakelse!

Helt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg?

Svaret ditt er ikke bindende.

Distribution

Distribution of values for fo_media

209 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
fo_media	Vi takker så mye for din deltakelse!

Helt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg?

Svaret ditt er ikke bindende. |haven_labelled | 209| 0.4951691|4 |4 |5 | 4.346342| 0.4769683| 2|▇▁▁▁▁▁▁▅ |F40.0 |5 |

Value labels

Response choices

name	value
Veldig negativ	4
Veldig negativ	5

on_sikkert_post

Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen?

Distribution

Distribution of values for on_sikkert_post

325 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
on_sikkert_post	Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen?	haven_labelled	325	0.2149758	1	3	3	2.52809	0.6757132	3	▁▁▁▃▁▁▁▇	F40.0	5

Value labels

Response choices

name	value
Veldig negativ	1
Veldig negativ	2
Veldig negativ	3

fg

1 == Treatment group (AI), 0 == control group (carbon capture)

Distribution

Distribution of values for fg

0 missing values.

Summary statistics

name	label	data_type	n_missing	complete_rate	min	median	max	mean	sd	hist
fg	1 == Treatment group (AI), 0 == control group (carbon capture)	numeric	0	1	0	1	1	0.5700483	0.4956679	▆▁▁▁▇

Missingness report

Codebook table

name	label	data_type	value_labels	n_missing	complete_rate	min	median	max	mean	sd	n_value_labels	hist	format.spss	display_width
sikker_fg	Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.	haven_labelled	1. Jeg vil delta i undersøkelsen, 2. Jeg samtykker ikke og vil ikke delta i undersøkelsen eller det påfølgende arrangementet	414	0.0000000	Inf	NA	-Inf	NaN	NA	2		F40.0	5
fg_kunnskap_1	First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence	haven_labelled	1. Very good knowledge, 2. Good knowledge, 3. Somwhat good knowledge, 4. Little knowledge, 5. No knowledge at all, 6. Do not know	1	0.9975845	1	4	6	3.6150121	0.7755157	6	▁▁▁▃▇▁▁▁	F40.0	5
fg_kunnskap_2	First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway	haven_labelled	1. Very good knowledge, 2. Good knowledge, 3. Somwhat good knowledge, 4. Little knowledge, 5. No knowledge at all, 6. Do not know	1	0.9975845	1	4	6	3.8837772	0.9507551	6	▁▂▁▂▇▁▃▁	F40.0	5
fg_kunnskap_3	First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work	haven_labelled	1. Very good knowledge, 2. Good knowledge, 3. Somwhat good knowledge, 4. Little knowledge, 5. No knowledge at all, 6. Do not know	1	0.9975845	1	4	6	3.8668281	0.9947380	6	▁▂▁▃▇▁▅▁	F40.0	5
fg_forbedring	Based on what you know, do you think the Norwegian authorities’ use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?	haven_labelled	1. Very strong improvement, 2. Strong improvement, 3. Some improvement, 4. Neither improvement or worsening, 5. Some worsening, 6. Strong worsening, 7. Very strong worsening, 9. Do not know	1	0.9975845	1	3	9	3.7167070	1.9594523	8	▃▇▂▂▁▁▁▂	F40.0	5
fg_likhet_1	And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it	haven_labelled	1. Very much more, 2. More, 3. Somwhat more, 4. Neither more or less, 6. Less, 7. Very much less, 8. Do not know	1	0.9975845	1	4	8	4.4721550	1.9514308	7	▁▃▇▅▅▃▂▃	F40.0	5
fg_likhet_2	And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector	haven_labelled	1. Very much more, 2. More, 3. Somwhat more, 4. Neither more or less, 5. Somewhat less, 6. Less, 7. Very much less, 8. Do not know	1	0.9975845	1	3	8	3.3365617	1.7265895	8	▂▆▇▃▁▁▁▂	F40.0	5
fg_likhet_3	And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities’ decisions	haven_labelled	1. Very much more, 2. More, 3. Somwhat more, 4. Neither more or less, 5. Somewhat less, 6. Less, 7. Very much less, 8. Do not know	1	0.9975845	1	4	8	4.5302663	1.6972791	8	▁▂▃▇▃▂▁▂	F40.0	5
fg_kunnskap2	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	1	0.9975845	1	5	8	4.6198547	1.3647052	8	▁▁▃▅▇▃▁▁	F40.0	5
fg_auto_flykt	Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 4. Do not know	1	0.9975845	1	1	4	1.2445521	0.6578507	3	▇▁▁▁▁▁▁▁	F40.0	5
fg_region	Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	1	0.9975845	1	5	8	4.6295400	1.5472491	8	▁▂▃▇▇▆▁▂	F40.0	5
fg_kjennskap_nav	Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	1	0.9975845	1	5	8	4.6077482	1.5768772	8	▁▂▃▂▇▃▁▁	F40.0	5
fg_auto_nav	Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	1	0.9975845	1	1	3	1.1961259	0.4702659	3	▇▁▁▁▁▁▁▁	F40.0	5
fg_pari_nav	Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?	haven_labelled	1. The first algorithm seems most fair, 2. The second algorithm seems most fair, 3. Do not know	1	0.9975845	1	1	3	1.5956416	0.7265294	3	▇▁▁▅▁▁▁▂	F40.0	5
fg_kjennskap_prv	Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?	haven_labelled	1. Oppose very strongly, 2. Oppose strongly, 3. Oppose somewhat, 4. Neither oppose or support, 5. Support somewhat, 6. Support strongly, 7. Support very stringly, 8. Do not know	1	0.9975845	1	4	8	3.9055690	1.7273620	8	▂▃▅▃▇▂▁▁	F40.0	5
fg_auto_prv	Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	1	0.9975845	1	1	3	1.1089588	0.3944298	3	▇▁▁▁▁▁▁▁	F40.0	5
fg_imp_1	How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees	haven_labelled	1. Not important at all, 2. Not important, 3. Somwhat important, 4. Very important, 5. Extremely improtant, 6. Do not know	1	0.9975845	1	4	6	3.9661017	0.7719686	6	▁▁▁▃▇▁▃▁	F40.0	5
fg_imp_2	How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting	haven_labelled	1. Not important at all, 2. Not important, 3. Somwhat important, 4. Very important, 5. Extremely improtant, 6. Do not know	1	0.9975845	1	4	6	3.8668281	0.7908388	6	▁▁▁▅▇▁▂▁	F40.0	5
fg_imp_3	How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates	haven_labelled	1. Not important at all, 2. Not important, 3. Somwhat important, 4. Very important, 5. Extremely improtant, 6. Do not know	1	0.9975845	1	5	6	4.4406780	0.7237252	6	▁▁▁▁▆▁▇▁	F40.0	5
fg_flykt_post	Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	1	0.9975845	1	2	3	1.7191283	0.5561197	3	▅▁▁▇▁▁▁▁	F40.0	5
fg_nav_post	Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	1	0.9975845	1	2	3	1.6222760	0.5934003	3	▇▁▁▇▁▁▁▁	F40.0	5
fg_prv_post	Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?	haven_labelled	1. Manual processing, 2. Algorithmic processing, 3. Do not know	1	0.9975845	1	2	3	1.6246973	0.6661513	3	▇▁▁▇▁▁▁▂	F40.0	5
fg_tradeoff_1	Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?	haven_labelled	1. A, 2. B, 3. Do not know	1	0.9975845	1	1	3	1.6973366	0.7961788	3	▇▁▁▅▁▁▁▃	F40.0	5
fg_tradeoff_2	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two	haven_labelled	1. X, 2. Y, 3. Do not know	1	0.9975845	1	2	3	1.9661017	0.8583194	3	▇▁▁▆▁▁▁▇	F40.0	5
fg_tradeoff_3	Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive	haven_labelled	1. One, 2. Two, 3. Do not know	2	0.9951691	1	2	3	1.9805825	0.7981802	3	▇▁▁▇▁▁▁▇	F40.0	5
demo_edu	What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.	haven_labelled	1. No eductation, 2. Pre-school, 3. High school, 5. Vocational school level (includes educations that are based on upper secondary school, but which are not approved as university and college education), 6. University or collage, up to 4 yearrs, 7. University or collage, more than 4 yearrs, 9. Other	207	0.5000000	2	6	9	5.5603865	1.5718939	7	▁▅▁▃▇▇▁▁	F40.0	5
demo_party	Which party would you vote for if there were a parliamentary election tomorrow?	haven_labelled	1. Progress party, 2. Conservative party, 3. Liberal Party, 4. Christian democrats, 13. Green party, 5. Center party, 6. Labour party, 7. Socialist party, 8. Red party, 9. Other, 10. Could not vote, 11. Would not vote	208	0.4975845	1	6	13	6.1456311	3.6836031	12	▇▃▁▇▃▂▂▃	F40.0	5
demo_trust	How much trust or distrust do you have in scientists?	haven_labelled	1. Very high trust, 2. High trust, 3. Some trust, 4. Neither trust nor mistrust, 5. Some mistrust, 6. High mistrust, 7. Very high mistrust, 8. Do not know	207	0.5000000	1	2	7	2.3864734	1.0727927	8	▂▇▂▂▁▁▁▁	F40.0	5
demo_lr_scale	In politics, you often talk about the ‘left’ and the ‘right’. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?	haven_labelled	0. 0 - Left, 1. 1, 2. 2, 3. 3, 4. 4, 5. 5, 6. 6, 7. 7, 8. 8, 9. 9, 10. 10, 11. 11- Right	208	0.4975845	0	5	11	5.1650485	2.7975498	12	▂▃▇▆▃▆▁▃	F40.0	5
post	0 == Pre deliberation survey responses, 1 == post delibiration survey	numeric	.	0	1.0000000	0	0.5	1	0.5000000	0.5006050	NA	▇▁▁▁▇	NA	NA
fg_sikkert_post	Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?	haven_labelled	1. Veldig negativ, 2. Veldig negativ, 3. Veldig negativ	297	0.2826087	1	3	3	2.6666667	0.5085476	3	▁▁▁▃▁▁▁▇	F40.0	5
fo_delib	Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?	haven_labelled	1. Veldig negativ, 2. Veldig negativ, 3. Veldig negativ, 4. Veldig negativ	208	0.4975845	1	2	3	1.6262136	0.6097518	4	▇▁▁▇▁▁▁▁	F40.0	5
fo_media	Vi takker så mye for din deltakelse!

Helt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg?

Svaret ditt er ikke bindende. |haven_labelled |4. Veldig negativ,
5. Veldig negativ | 209| 0.4951691|4 |4 |5 | 4.3463415| 0.4769683| 2|▇▁▁▁▁▁▁▅ |F40.0 |5 | |on_sikkert_post |Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen? |haven_labelled |1. Veldig negativ,
2. Veldig negativ,
3. Veldig negativ | 325| 0.2149758|1 |3 |3 | 2.5280899| 0.6757132| 3|▁▁▁▃▁▁▁▇ |F40.0 |5 | |fg |1 == Treatment group (AI), 0 == control group (carbon capture) |numeric |. | 0| 1.0000000|0 |1.0 |1 | 0.5700483| 0.4956679| NA|▆▁▁▁▇ |NA |NA |

JSON-LD metadata

The following JSON-LD can be found by search engines, if you share this codebook publicly on the web.

{
  "name": ".",
  "datePublished": "2022-11-16",
  "description": "The dataset has N=414 rows and 35 columns.\n0 rows have no missing values on any column.\n\n\n## Table of variables\nThis table contains variable names, labels, and number of missing values.\nSee the complete codebook for more.\n\n[truncated]\n\n### Note\nThis dataset was automatically described using the [codebook R package](https://rubenarslan.github.io/codebook/) (version 0.9.2).",
  "keywords": ["sikker_fg", "fg_kunnskap_1", "fg_kunnskap_2", "fg_kunnskap_3", "fg_forbedring", "fg_likhet_1", "fg_likhet_2", "fg_likhet_3", "fg_kunnskap2", "fg_auto_flykt", "fg_region", "fg_kjennskap_nav", "fg_auto_nav", "fg_pari_nav", "fg_kjennskap_prv", "fg_auto_prv", "fg_imp_1", "fg_imp_2", "fg_imp_3", "fg_flykt_post", "fg_nav_post", "fg_prv_post", "fg_tradeoff_1", "fg_tradeoff_2", "fg_tradeoff_3", "demo_edu", "demo_party", "demo_trust", "demo_lr_scale", "post", "fg_sikkert_post", "fo_delib", "fo_media", "on_sikkert_post", "fg"],
  "@context": "http://schema.org/",
  "@type": "Dataset",
  "variableMeasured": [
    {
      "name": "sikker_fg",
      "description": "Are you sure that you do not consent to our data processing and want to end the survey? You cannot then participate in the event on 11 June either, as the consent also applies to that event.",
      "value": "1. Jeg vil delta i undersøkelsen,\n2. Jeg samtykker ikke og vil ikke delta i undersøkelsen eller det påfølgende arrangementet",
      "maxValue": 2,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap_1",
      "description": "First, how well would you say you have the following knowledge: - Machine learning and artificial intelligence",
      "value": "1. Very good knowledge,\n2. Good knowledge,\n3. Somwhat good knowledge,\n4. Little knowledge,\n5. No knowledge at all,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap_2",
      "description": "First, how well would you say you know the following: - How the Norwegian authorities currently decide where refugees will be settled once they have been granted residence in Norway",
      "value": "1. Very good knowledge,\n2. Good knowledge,\n3. Somwhat good knowledge,\n4. Little knowledge,\n5. No knowledge at all,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap_3",
      "description": "First, how well would you say you know the following: - How NAV currently decides which people on sick leave should be invited to a dialogue meeting about how they can more easily return to work",
      "value": "1. Very good knowledge,\n2. Good knowledge,\n3. Somwhat good knowledge,\n4. Little knowledge,\n5. No knowledge at all,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_forbedring",
      "description": "Based on what you know, do you think the Norwegian authorities' use of machine learning and artificial intelligence will lead to an improvement or worsening of public services?",
      "value": "1. Very strong improvement,\n2. Strong improvement,\n3. Some improvement,\n4. Neither improvement or worsening,\n5. Some worsening,\n6. Strong worsening,\n7. Very strong worsening,\n9. Do not know",
      "maxValue": 9,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_likhet_1",
      "description": "And what do you think it will lead to in terms of the following points? - Verifiability: Possibility of getting an explanation of the decision for those who are affected by it",
      "value": "1. Very much more,\n2. More,\n3. Somwhat more,\n4. Neither more or less,\n6. Less,\n7. Very much less,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_likhet_2",
      "description": "And what do you think it will lead to in terms of the following points? - Impartiality: Possibility for citizens to be treated equally by the public sector",
      "value": "1. Very much more,\n2. More,\n3. Somwhat more,\n4. Neither more or less,\n5. Somewhat less,\n6. Less,\n7. Very much less,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_likhet_3",
      "description": "And what do you think it will lead to in terms of the following points? - Legitimacy: Likelihood that the citizens will voluntarily follow the authorities' decisions",
      "value": "1. Very much more,\n2. More,\n3. Somwhat more,\n4. Neither more or less,\n5. Somewhat less,\n6. Less,\n7. Very much less,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kunnskap2",
      "description": "Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of algorithms based on machine learning and artificial intelligence when settling refugees?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_auto_flykt",
      "description": "Imagine two situations. In one situation, a case manager will settle refugees based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved in the case (fully automated procedure). Which of these two procedures do you prefer?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n4. Do not know",
      "maxValue": 4,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_region",
      "description": "Algorithms that are developed based on machine learning and artificial intelligence need large amounts of data to be able to create accurate models for the settlement of refugees. It then becomes difficult to create algorithms that are accurate enough for each individual municipality, since there is too little basis on which to base the model on. An alternative is to merge larger areas into labor market regions, and request the settlement of refugees according to these regions. How strongly do you support or oppose requests for the settlement of refugees being made by region instead of municipalities?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kjennskap_nav",
      "description": "Based on your knowledge, to what extent do you support that NAV should be open to using artificial intelligence when deciding who should be invited to a dialogue meeting?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_auto_nav",
      "description": "Imagine two situations. In one situation, a case manager will invite to a dialogue meeting based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_pari_nav",
      "description": "Machine learning involves getting computers to learn to solve tasks based on data material. A dilemma that often arises is that different algorithms have different advantages and disadvantages, and you have to choose which considerations to prioritize. Imagine that NAV has to choose between using one of these two algorithms: The first algorithm is most accurate. Among those who need a dialogue meeting, however, more [Field-fairgov_pari_treat] are offered a dialogue meeting, even though there are an equal number of women and men who are on sick leave and need a dialogue meeting. The proportion who need a dialogue meeting without receiving an offer is therefore larger at [Field-fairgov_pari_treat_2]. The second algorithm ensures that the proportion of those on sick leave who are called to a dialogue meeting is equal for women and men. However, it is less accurate overall, so that fewer people who need a dialogue meeting are called. This applies to both women and men. If it was only between these two, which one do you think seems more fair?",
      "value": "1. The first algorithm seems most fair,\n2. The second algorithm seems most fair,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_kjennskap_prv",
      "description": "Based on your knowledge, how strongly do you support or oppose the Norwegian authorities opening up the use of artificial intelligence when deciding who should be paroled from prison?",
      "value": "1. Oppose very strongly,\n2. Oppose strongly,\n3. Oppose somewhat,\n4. Neither oppose or support,\n5. Support somewhat,\n6. Support strongly,\n7. Support very stringly,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_auto_prv",
      "description": "Imagine two situations. In one situation, a case manager will grant a prison inmate parole based on his own assessment with the support of an algorithm (semi-automated procedure). In the second situation, it is the same algorithm that makes the decision completely alone, without a case manager being involved (fully automated procedure). Which of these two procedures do you prefer?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_imp_1",
      "description": "How important do you think it is that the authorities come to the right decisions in the following matters: - Settlement of refugees",
      "value": "1. Not important at all,\n2. Not important,\n3. Somwhat important,\n4. Very important,\n5. Extremely improtant,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_imp_2",
      "description": "How important do you think it is that the authorities come to the right decisions in the following matters: - Calling for a dialogue meeting",
      "value": "1. Not important at all,\n2. Not important,\n3. Somwhat important,\n4. Very important,\n5. Extremely improtant,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_imp_3",
      "description": "How important do you think it is that the authorities come to the right decisions in the following cases: - Parole of prison inmates",
      "value": "1. Not important at all,\n2. Not important,\n3. Somwhat important,\n4. Very important,\n5. Extremely improtant,\n6. Do not know",
      "maxValue": 6,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_flykt_post",
      "description": "Case 1 of 3: Settlement of refugees As a refugee to Norway, you have been granted asylum, and are waiting in an asylum reception center to be settled in a municipality. You get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_bo_treat] will take much more time than the algorithm, which will extend your stay at the asylum reception accordingly. Which decision-making process would you have chosen?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_nav_post",
      "description": "Case 2 of 3: Dialogue meeting with NAV As sick leave, you receive follow-up from NAV. They will assess whether you should be invited to a dialogue meeting to discuss what can be done to enable you to return to work. Dialogue meetings require resources, so not everyone can get that offer. When NAV assesses your need, you get the choice between having your case assessed manually by a case manager or fully automatically using an algorithm. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_di_treat] will take much more time than the algorithm, which extends the period accordingly before you possibly receive an invitation to a dialogue meeting. Which decision-making process would you have chosen?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_prv_post",
      "description": "Case 3 of 3: Parole As an inmate, you have applied for parole, and are given the choice between having your case assessed manually or automatically. You can expect the case manager to take longer than the algorithm to assess your case. Normally, [Field-fairgov_manvsmach_pr_treat] will take much more time than the algorithm, which extends the period accordingly before you can possibly be paroled. Which decision-making process would you have chosen?",
      "value": "1. Manual processing,\n2. Algorithmic processing,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_tradeoff_1",
      "description": "Imagine that you have to choose between machine learning models A and B that differ by the characteristics shown in the figure below. The characteristics of Model A are shown by the two columns on the left, and the characteristics of Model B are shown by the two columns on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole. Model A has a lower proportion of wrong decisions (33%) than Model B (66%). Model B has a lower proportion of incorrect rejections (33%) than Model A (66%). Which of these two do you perceive to be fairer?",
      "value": "1. A,\n2. B,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_tradeoff_2",
      "description": "Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models X and Y that differ in the characteristics shown in the figure below. The features of Model X are shown on the left and the features of Model Y are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.False denial rate: How often the model is expected to deny parole to someone who should actually be granted parole.Model X misses equally for black and white Americans, but falsely rejects a greater share of whites (45%) than blacks (24%).Model Y falsely rejects an equal share of white and black Americans, but misses more whites (47%) than blacks (21%). Which of these two",
      "value": "1. X,\n2. Y,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg_tradeoff_3",
      "description": "Algorithm-based parole has been used in the United States. Some believe that this difference between blacks and whites should be equalized in their chance of being paroled, while others believe that everyone should receive the same process, regardless of origin.Imagine that you have to choose between machine learning models 1 and 2 which differ in the characteristics shown in the figure below. The characteristics of Model 1 are shown on the left and the characteristics of Model 2 are shown on the right. Rate of wrong decisions: How often the model is expected to miss a decision. A lower number means it is more accurate.Proportion that gets parole: What proportion will get parole from the model. Model 1 has an equal share of wrongful convictions between black and white Americans, but grants parole to more whites (55%) than blacks (34%). Model 2 has a greater proportion of wrongful convictions for white (47%) than black (21%) Americans, but grants parole to an equal proportion in each group. Which of these two do you perceive",
      "value": "1. One,\n2. Two,\n3. Do not know",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_edu",
      "description": "What is your highest completed education? If you have several educations at the same level, choose the one that is most relevant to you.",
      "value": "1. No eductation,\n2. Pre-school,\n3. High school,\n5. Vocational school level (includes educations that are based on upper secondary school, but which are not approved as university and college education),\n6. University or collage, up to 4 yearrs,\n7. University or collage, more than 4 yearrs,\n9. Other",
      "maxValue": 9,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_party",
      "description": "Which party would you vote for if there were a parliamentary election tomorrow?",
      "value": "1. Progress party,\n2. Conservative party,\n3. Liberal Party,\n4. Christian democrats,\n13. Green party,\n5. Center party,\n6. Labour party,\n7. Socialist party,\n8. Red party,\n9. Other,\n10. Could not vote,\n11. Would not vote",
      "maxValue": 13,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_trust",
      "description": "How much trust or distrust do you have in scientists?",
      "value": "1. Very high trust,\n2. High trust,\n3. Some trust,\n4. Neither trust nor mistrust,\n5. Some mistrust,\n6. High mistrust,\n7. Very high mistrust,\n8. Do not know",
      "maxValue": 8,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "demo_lr_scale",
      "description": "In politics, you often talk about the 'left' and the 'right'. Below is a scale where 0 represents those who are on the far left politically, and 10 represents those who are on the far right politically. How would you place yourself on such a scale?",
      "value": "0. 0 - Left,\n1. 1,\n2. 2,\n3. 3,\n4. 4,\n5. 5,\n6. 6,\n7. 7,\n8. 8,\n9. 9,\n10. 10,\n11. 11- Right",
      "maxValue": 11,
      "minValue": 0,
      "@type": "propertyValue"
    },
    {
      "name": "post",
      "description": "0 == Pre deliberation survey responses, 1 == post delibiration survey",
      "value": ". ",
      "maxValue": "-Inf",
      "minValue": "Inf",
      "@type": "propertyValue"
    },
    {
      "name": "fg_sikkert_post",
      "description": "Do you feel more confident or more uncertain in your view of the use of artificial intelligence in administration after you participated in the deliberation?",
      "value": "1. Veldig negativ,\n2. Veldig negativ,\n3. Veldig negativ",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fo_delib",
      "description": "Now that you have participated in this deliberation, we want to know what you think about including ordinary citizens in political discussions on this kind of topic. Do you think that ordinary people have a lot of value to contribute and that politicians and experts should listen to this, or do you think that ordinary people have nothing valuable to contribute and that the discussions should be left to politicians and experts?",
      "value": "1. Veldig negativ,\n2. Veldig negativ,\n3. Veldig negativ,\n4. Veldig negativ",
      "maxValue": 4,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fo_media",
      "description": "Vi takker så mye for din deltakelse!  \r\n\r\nHelt til slutt: Det kan hende noen media ønsker å komme i kontakt med enkelte deltakere for å høre om deres opplevelser med å delta i dette arrangementet. Dersom det skulle komme en slik henvendelse, vil du være villig til at de tar kontakt med deg? \r\n\r\nSvaret ditt er ikke bindende.",
      "value": "4. Veldig negativ,\n5. Veldig negativ",
      "maxValue": 5,
      "minValue": 4,
      "@type": "propertyValue"
    },
    {
      "name": "on_sikkert_post",
      "description": "Føler du deg mer sikker eller mer usikker i ditt syn på ulike metoder for å fjerne CO2 fra luften etter at du deltok i deliberasjonen?",
      "value": "1. Veldig negativ,\n2. Veldig negativ,\n3. Veldig negativ",
      "maxValue": 3,
      "minValue": 1,
      "@type": "propertyValue"
    },
    {
      "name": "fg",
      "description": "1 == Treatment group (AI), 0 == control group (carbon capture)",
      "value": ". ",
      "maxValue": "-Inf",
      "minValue": "Inf",
      "@type": "propertyValue"
    }
  ]
}`

Socio-demographics